The following handle holds various files of this Leiden University dissertation:
http://hdl.handle.net/1887/71806
Author: Son, D. van
Title: EEG theta/beta ratio: a marker of executive control and its relation with
anxiety-linked attentional bias for threat
CHAPTER 5.
ABSTRACT
Background: a growing number of studies suggest that EEG theta/beta ratio (TBR) is inversely related to executive cognitive control. Neurofeedback training aimed at reducing TBR (TBR NFT) might provide a tool to study causality in this relation and might enhance human performance.
Goal: to investigate whether TBR NFT lowers TBR in healthy participants
Method: Twelve healthy female participants were assigned (single blind) to one of three groups. Groups differed on baseline durations and one group received only sham NFT. TBR NFT consisted of eight or fourteen 25-minutes sessions.
Results: No evidence was found that TBR NFT had any effect on TBR.
Conclusions: The current TBR NFT protocol is ineffective. This replicates a previous study with a different protocol. TBR NFT may not be effective in healthy participants.
Resting state encephalographic (EEG) signals are composed of different frequency components, many of which are found to be relatively stable over time (Williams, Simms, Clark, Paul, Rowe & Gordon, 2005). Specific spectrum components reflect functional neural activity as an electrophysiological correlate with certain behaviors (Hofman & Schutter, 2012; Sutton & Davidson, 2000). For example, the ratio between activity in the theta band (4-7 Hz) and activity in the beta band (13-30 Hz), called theta/beta ratio (TBR), has been related to different aspects of cognitive control and motivated decision making (Schutter & van Honk 2005a; Massar, Rossi, Schutter & Kenemans, 2012; Massar, Kenemans & Schutter, 2014), to attentional control in healthy young adults (Putman, van Peer, Maimari & van der Werff, 2010; Putman, Verkuil, Arias-Garcia, Pantazi & van Schie, 2014; Angelidis, Schakel, van der Does & Putman., 2016) and to reversal learning (Wischnewski, Zerr & Schutter, 2016). Additionally, a higher baseline TBR was found to correlate to a stronger decline in cognitive control after stress-induction (Putman et al., 2014). TBR has a high test-retest reliability and predicts attentional control scores over a one-week interval (Angelidis et al., 2016). All in all, TBR is likely a stable electrophysiological marker of executive control.
Attentional control refers to the capacity to efficiently choose where to pay attention to (Posner & Petersen, 1990). When attentional control fails, attentional bias towards threat can be the result. This may happen when stress or anxiety prioritizes the processing of mildly threatening distracters. In other words, during anxious states, bottom-up processing of threatening distracters is facilitated, while top-down executive functions are inhibited (Derakshan & Eysenck, 2009). Furthermore, executive control can be decreased by distracting thoughts that accompany stress or anxiety, impairing working memory (Coy, O'Brien, Tabaczynski, Northern, Carels, 2011; Eysenck, Derakshan, Santos, Calvo, 2007). This is in line with the widely accepted idea that test anxiety causes divided attention, leading to for example lower academic performance (Hembree, 1988; Duty Christian , Loftus, Zappi, 2016).
TBR was found to be related to trait attentional control (Putman et al., 2010; Putman et al., 2014; Angelidis, Schakel & van der Does & Putman, 2016), to resilience to the effects of stress on task performance (Putman et al., 2014), to down regulation of negative affect (Tortella-Feliu , Morillas-Romero, Balle, Llabrés, Bornas, Putman, 2014) and to regulation of automatic attentional bias to threat; (Angelidis, Hagenaars, van der Does, van Son & Putman, 2018). The study of TBR is therefore potentially interesting for a range of phenomena, conditions and applications, such as stress-cognition interactions, anxious psychopathology or human performance enhancement. Experimentally manipulating TBR could give further insights in causal relations between this EEG marker, cognitive control and stress effects, as well as possibly pave the way for future development of interventions.
5
ABSTRACT
Background: a growing number of studies suggest that EEG theta/beta ratio (TBR) is inversely related to executive cognitive control. Neurofeedback training aimed at reducing TBR (TBR NFT) might provide a tool to study causality in this relation and might enhance human performance.
Goal: to investigate whether TBR NFT lowers TBR in healthy participants
Method: Twelve healthy female participants were assigned (single blind) to one of three groups. Groups differed on baseline durations and one group received only sham NFT. TBR NFT consisted of eight or fourteen 25-minutes sessions.
Results: No evidence was found that TBR NFT had any effect on TBR.
Conclusions: The current TBR NFT protocol is ineffective. This replicates a previous study with a different protocol. TBR NFT may not be effective in healthy participants.
Resting state encephalographic (EEG) signals are composed of different frequency components, many of which are found to be relatively stable over time (Williams, Simms, Clark, Paul, Rowe & Gordon, 2005). Specific spectrum components reflect functional neural activity as an electrophysiological correlate with certain behaviors (Hofman & Schutter, 2012; Sutton & Davidson, 2000). For example, the ratio between activity in the theta band (4-7 Hz) and activity in the beta band (13-30 Hz), called theta/beta ratio (TBR), has been related to different aspects of cognitive control and motivated decision making (Schutter & van Honk 2005a; Massar, Rossi, Schutter & Kenemans, 2012; Massar, Kenemans & Schutter, 2014), to attentional control in healthy young adults (Putman, van Peer, Maimari & van der Werff, 2010; Putman, Verkuil, Arias-Garcia, Pantazi & van Schie, 2014; Angelidis, Schakel, van der Does & Putman., 2016) and to reversal learning (Wischnewski, Zerr & Schutter, 2016). Additionally, a higher baseline TBR was found to correlate to a stronger decline in cognitive control after stress-induction (Putman et al., 2014). TBR has a high test-retest reliability and predicts attentional control scores over a one-week interval (Angelidis et al., 2016). All in all, TBR is likely a stable electrophysiological marker of executive control.
Attentional control refers to the capacity to efficiently choose where to pay attention to (Posner & Petersen, 1990). When attentional control fails, attentional bias towards threat can be the result. This may happen when stress or anxiety prioritizes the processing of mildly threatening distracters. In other words, during anxious states, bottom-up processing of threatening distracters is facilitated, while top-down executive functions are inhibited (Derakshan & Eysenck, 2009). Furthermore, executive control can be decreased by distracting thoughts that accompany stress or anxiety, impairing working memory (Coy, O'Brien, Tabaczynski, Northern, Carels, 2011; Eysenck, Derakshan, Santos, Calvo, 2007). This is in line with the widely accepted idea that test anxiety causes divided attention, leading to for example lower academic performance (Hembree, 1988; Duty Christian , Loftus, Zappi, 2016).
TBR was found to be related to trait attentional control (Putman et al., 2010; Putman et al., 2014; Angelidis, Schakel & van der Does & Putman, 2016), to resilience to the effects of stress on task performance (Putman et al., 2014), to down regulation of negative affect (Tortella-Feliu , Morillas-Romero, Balle, Llabrés, Bornas, Putman, 2014) and to regulation of automatic attentional bias to threat; (Angelidis, Hagenaars, van der Does, van Son & Putman, 2018). The study of TBR is therefore potentially interesting for a range of phenomena, conditions and applications, such as stress-cognition interactions, anxious psychopathology or human performance enhancement. Experimentally manipulating TBR could give further insights in causal relations between this EEG marker, cognitive control and stress effects, as well as possibly pave the way for future development of interventions.
situations with counterproductive interactions between stress and cognition, such as music performance under stressful conditions (Egner & Gruzelier, 2003). NFT has also been applied for reducing symptoms of Attention Deficit Hyperactivity Disorder (ADHD). ADHD has very often been associated with high TBR (for review and meta-analysis, see Arns et al., 2013; Barry, Clarke & Johnstone, 2003) and NFT targeting TBR has been found to successfully reduce TBR and ADHD-related symptoms in individuals diagnosed with ADHD (e.g. hyperactivity, impaired attention; e.g. Butnik, 2005; Leins et al., 2007; Lubar et al., 1995; Kouijzer, de Moor, Gerrits, Congedo and van Schie, 2009; Janssen, Bink, Weeda, Geladé, van Mourik, Maras & Oosterlaan, 2016; for a review see Vernon, 2005).
The study of the potential beneficial effects of TBR-reducing NFT seems warranted given the
abovementioned relations between TBR and various psychological regulatory constructs. However, because we believe it is imperative to first ascertain that indeed reliable changes in TBR by NFT can be observed, we selected healthy participants with mildly elevated TBR. We aimed to investigate whether NFT induces changes in TBR in people with mildly elevated TBR but who do not have a clinical diagnosis of psychopathology. The primary outcome measure of this study was changes in the targeted EEG parameters. These changes are likely easier to detect and more consistent than changes at the more multifaceted and complex behavioral level.
Doppelmayr and Weber (2011) previously investigated whether a TBR NFT protocol exerts the intended effects on the EEG spectrum level in healthy individuals, not selected on TBR level. The effect of the NFT TBR training on its trained EEG indices was compared to the effect of an NFT protocol training the ‘Sensori- Motor Rhythms’ (SMR; 12-15 Hz) and a sham-NFT with daily changing frequency bands. Healthy individuals who received the SMR training protocol were able to significantly modulate their EEG in the trained frequency band, whereas individuals who received the TBR or sham protocol were not. To our knowledge, this is the only study to date that directly investigated TBR NFT in healthy individuals by primarily looking at the direct effects on the EEG parameters of interest. We aimed to replicate and extend Doppelmayr and Weber’s findings by testing in an independent study again if TBR can be changed.
Our hypothesis that a TBR NFT can induce changes in EEG for individuals with mildly elevated TBR has not been studied extensively yet. Subjecting participants from this population to a very lengthy active NFT TBR training is demanding on the participants and could potentially cause unknown side effects. The best approach would be to study a small sample in depth, by thoroughly inspecting effects of active-NFT in each individual per session. We therefore employed a multiple baseline case series design. This design was chosen in order to closely examine any possible change in TBR at the level of the studied individuals so as not to overlook possible leads to increase NFT effectiveness and to minimize the chance of prematurely ruling out potential effectiveness of NFT for our purposes. A multiple baseline case series design involves the measurement of multiple persons both before and after an intervention (Watson & Workman, 1981). In this design the start of the intervention is varied sequentially across individuals or small groups of individuals. During the baseline phase before intervention, the behavior or measure of interest is measured a number of times to observe its natural variation over time. When a change only takes place shortly after a specific intervention is introduced and not following a different
intervention, the change can be attributed to the intervention (Baer et al., 1968; Kinugasa et al., 2004; Koehler &
Levin, 2000). A frequently used method in a case series design is visual inspection, which provides a reliable alternative for statistical tests for detecting changes by intervention, when sample sizes are too small for good statistical power (Fisher et al., 2003).
There are other, mainly ethical benefits to the smaller number of sessions required for a case series design. First of all, executing a controlled study with large sample sizes (see Cohen et al., 2014) places a lot of burden on the test-participants and implies a big investment of societal resources that may not be warranted yet. Additionally, nothing is known about possible negative side-effects in our intended population although the literature does suggest that such effects might exist. Low TBR for example has been related to low approach-driven or hedonically motivated behavior as measured with the IOWA gambling task (which has been associated with depression and anxiety; Cella, Dymond & Cooper, 2010; Massar et al., 2012; Massar et al., 2014; Mueller, Nquyen, Ray, Borkovec, 2009; Schutter & van Honk, 2005) and as measured with the self-report BIS/BAS scale (Putman et al., 2010; Carver & White, 1994). Also, two studies (Putman et al., 2010; Angelidis et al., 2016)
demonstrate a negative association between TBR and self-reported negative, anxious affect as measured with the State-Trait Anxiety Inventory (STAI-State; Spielberger, 1983; van der Ploeg, Defares & Spielberger, 1980). Finally, one study (Enriquez-Geppert et al., 2014) has reported beneficial effects of working memory performance of a theta-only upregulation using NFT in healthy participants. All in all, at this stage of TBR NFT research in healthy adults, where side effects are not yet thoroughly investigated but cannot be ruled out, applying this intervention in a large group of participants and over a long time period is not yet defensible.
We assessed whether NFT reduces TBR in healthy individuals with mildly elevated TBR. TBR was the primary outcome; self-reported attentional control and state anxiety were assessed as secondary outcomes and to measure potentially unwanted side effects of NFT. A multiple baseline design was used employing various durations of baseline, sham-NFT and active-NFT sessions. We expected to see a reduction of TBR sometime after switching from only sessions to active treatment. We expected an absence of such measurement-only-controlled changes in a third sham-only group and, finally, we expected that TBR during the final sessions would be clearly lower in the two active NFT groups than in the sham-only group. Our primary interest was changes in TBR within the training sessions but we also looked at changes in TBR during resting state
measurements at the start of the sessions (between-sessions changes). Finaly, we performed in-depth exploration of the time course of TBR within and between training sessions, exploiting the case series’ benefits of temporally fine grained observation.
Methods Participants
5
situations with counterproductive interactions between stress and cognition, such as music performance under stressful conditions (Egner & Gruzelier, 2003). NFT has also been applied for reducing symptoms of Attention Deficit Hyperactivity Disorder (ADHD). ADHD has very often been associated with high TBR (for review and meta-analysis, see Arns et al., 2013; Barry, Clarke & Johnstone, 2003) and NFT targeting TBR has been found to successfully reduce TBR and ADHD-related symptoms in individuals diagnosed with ADHD (e.g. hyperactivity, impaired attention; e.g. Butnik, 2005; Leins et al., 2007; Lubar et al., 1995; Kouijzer, de Moor, Gerrits, Congedo and van Schie, 2009; Janssen, Bink, Weeda, Geladé, van Mourik, Maras & Oosterlaan, 2016; for a review see Vernon, 2005).
The study of the potential beneficial effects of TBR-reducing NFT seems warranted given the
abovementioned relations between TBR and various psychological regulatory constructs. However, because we believe it is imperative to first ascertain that indeed reliable changes in TBR by NFT can be observed, we selected healthy participants with mildly elevated TBR. We aimed to investigate whether NFT induces changes in TBR in people with mildly elevated TBR but who do not have a clinical diagnosis of psychopathology. The primary outcome measure of this study was changes in the targeted EEG parameters. These changes are likely easier to detect and more consistent than changes at the more multifaceted and complex behavioral level.
Doppelmayr and Weber (2011) previously investigated whether a TBR NFT protocol exerts the intended effects on the EEG spectrum level in healthy individuals, not selected on TBR level. The effect of the NFT TBR training on its trained EEG indices was compared to the effect of an NFT protocol training the ‘Sensori- Motor Rhythms’ (SMR; 12-15 Hz) and a sham-NFT with daily changing frequency bands. Healthy individuals who received the SMR training protocol were able to significantly modulate their EEG in the trained frequency band, whereas individuals who received the TBR or sham protocol were not. To our knowledge, this is the only study to date that directly investigated TBR NFT in healthy individuals by primarily looking at the direct effects on the EEG parameters of interest. We aimed to replicate and extend Doppelmayr and Weber’s findings by testing in an independent study again if TBR can be changed.
Our hypothesis that a TBR NFT can induce changes in EEG for individuals with mildly elevated TBR has not been studied extensively yet. Subjecting participants from this population to a very lengthy active NFT TBR training is demanding on the participants and could potentially cause unknown side effects. The best approach would be to study a small sample in depth, by thoroughly inspecting effects of active-NFT in each individual per session. We therefore employed a multiple baseline case series design. This design was chosen in order to closely examine any possible change in TBR at the level of the studied individuals so as not to overlook possible leads to increase NFT effectiveness and to minimize the chance of prematurely ruling out potential effectiveness of NFT for our purposes. A multiple baseline case series design involves the measurement of multiple persons both before and after an intervention (Watson & Workman, 1981). In this design the start of the intervention is varied sequentially across individuals or small groups of individuals. During the baseline phase before intervention, the behavior or measure of interest is measured a number of times to observe its natural variation over time. When a change only takes place shortly after a specific intervention is introduced and not following a different
intervention, the change can be attributed to the intervention (Baer et al., 1968; Kinugasa et al., 2004; Koehler &
Levin, 2000). A frequently used method in a case series design is visual inspection, which provides a reliable alternative for statistical tests for detecting changes by intervention, when sample sizes are too small for good statistical power (Fisher et al., 2003).
There are other, mainly ethical benefits to the smaller number of sessions required for a case series design. First of all, executing a controlled study with large sample sizes (see Cohen et al., 2014) places a lot of burden on the test-participants and implies a big investment of societal resources that may not be warranted yet. Additionally, nothing is known about possible negative side-effects in our intended population although the literature does suggest that such effects might exist. Low TBR for example has been related to low approach-driven or hedonically motivated behavior as measured with the IOWA gambling task (which has been associated with depression and anxiety; Cella, Dymond & Cooper, 2010; Massar et al., 2012; Massar et al., 2014; Mueller, Nquyen, Ray, Borkovec, 2009; Schutter & van Honk, 2005) and as measured with the self-report BIS/BAS scale (Putman et al., 2010; Carver & White, 1994). Also, two studies (Putman et al., 2010; Angelidis et al., 2016)
demonstrate a negative association between TBR and self-reported negative, anxious affect as measured with the State-Trait Anxiety Inventory (STAI-State; Spielberger, 1983; van der Ploeg, Defares & Spielberger, 1980). Finally, one study (Enriquez-Geppert et al., 2014) has reported beneficial effects of working memory performance of a theta-only upregulation using NFT in healthy participants. All in all, at this stage of TBR NFT research in healthy adults, where side effects are not yet thoroughly investigated but cannot be ruled out, applying this intervention in a large group of participants and over a long time period is not yet defensible.
We assessed whether NFT reduces TBR in healthy individuals with mildly elevated TBR. TBR was the primary outcome; self-reported attentional control and state anxiety were assessed as secondary outcomes and to measure potentially unwanted side effects of NFT. A multiple baseline design was used employing various durations of baseline, sham-NFT and active-NFT sessions. We expected to see a reduction of TBR sometime after switching from only sessions to active treatment. We expected an absence of such measurement-only-controlled changes in a third sham-only group and, finally, we expected that TBR during the final sessions would be clearly lower in the two active NFT groups than in the sham-only group. Our primary interest was changes in TBR within the training sessions but we also looked at changes in TBR during resting state
measurements at the start of the sessions (between-sessions changes). Finaly, we performed in-depth exploration of the time course of TBR within and between training sessions, exploiting the case series’ benefits of temporally fine grained observation.
Methods Participants
study (N = 54). Frontal TBR was chosen as preselection outcome variable since all previous studies had frontal TBR as main outcome variable and this is highly correlated with central TBR (r = 0.902, p<0.001 in the current study). We invited participants with the highest frontal TBR for the current study. Other inclusion criteria were: age between 18 and 24 years; no history of neurologic or psychiatric disorders; no history or current use of drugs other than low to moderate alcohol use or nicotine use and no use of medication that is known to directly influence the central nervous system. Recruitment took place at Leiden University, The Netherlands, between December 2015 and February 2016. All participants signed an informed consent and were free to terminate their participation at any time. For monetary compensation we used an incremental pay-off scheme, including disproportionately larger rewards for longer participation. This pay-off scheme was applied to minimize drop-out from the study. The study was approved by the local ethics review board (CEP16-011413), and pre-registered at ClinicalTrials.gov (NCT02763618).
Design
A single-blind case series multiple baseline design was used with a baseline (measurement-only period) varying prior to training onset and after training offset (Figure 5.1). Before the first lab session, the participants were assigned to one of three groups. Care was taken to obtain a more or less equal distribution of frontal TBR levels and age across the groups, but other than that the allocation to one of the three study groups was arbitrary. Allocation was done by the principal investigator who was not involved in the actual testing of the participants and had no contact with them. The experimenters that performed the study were kept blind to this allocation. All participants started with a three-session measurement-only phase with only a resting state EEG measurement. Participants in Group A continued with 14 sessions of active-NFT. Group B received six extra measurement-only sessions before they continued with eight active-NFT sessions. Group C received 14 sham-NFT sessions after the three-session measurement-only phase. A minimum of eight sessions active-NFT was applied because changes were usually seen around five or six 30-minute sessions in studies that found effects on theta frequency (Kao et al., 2014; Enriquez-Geppert et al., 2014). All participants were blind to which group they were in but the
experimenters were not blind to this for reasons of practical feasibility. During all sessions, questionnaires for state anxiety and state attentional control were assessed before every EEG measurement, active-NFT or sham-NFT. Our primary outcome variable was changes in frontal TBR within each session while our secondary outcome measurement was changes in frontal TBR between the sessions.
Figure 5.1. Difference in session course per group. Materials
Self-Report Questionnaires.During the first and last session, participants completed the trait version of the State-Trait Anxiety Inventory (STAI-Trait; Spielberger, 1983; van der Ploeg, Defares & Spielberger, 1980) and the Attentional Control Scale (ACS; Derryberry & Reed, 2002). The STAI-Trait assesses trait anxiety (20 items, range 20-80; Cronbach’s alpha = 0.89) and participants had to indicate their agreement with items like ‘I feel satisfied with myself’ and ‘I am a steady person’ on a four-point Likert scale. The ACS assesses self-reported attentional control in terms of attentional inhibition, attentional focus and the capacity to generate new thoughts (20 items, range 20-80; Cronbach’s alpha in present study = 0.85), e.g. ‘I can quickly switch from one task to another’. Self-reported state anxiety and state attentional control were measured on every session using the state version of the State-Trait Anxiety Inventory (STAI-State; Spielberger, 1983; van der Ploeg, Defares & Spielberger, 1980) and the state-Attentional Control questionnaire (s-AC; Angelidis & Putman, manuscript in preparation). STAI- State measures state anxiety at the moment of participation (20 items, range 20-80, Cronbach’s alpha = 0.91) and includes items like ‘I am tense’. s-AC measures attentional control at the moment of participation (6 items) and included items like ‘I feel very focused’ (Cronbach’s alpha = 0.84). The Behavioral Activation Scale (BAS; part of the Behavioral Inhibition and Activation Scale; BIS/BAS Carver & White, 1994) was assessed for the personality trait of behavioral activation (Cronbach’s alpha = 0.78). The BAS consists of the subscales BAS Reward, BAS Drive and BAS Fun Seeking but we assessed the total BAS score. STAI-t, ACS and BAS were included only to see if their scores changed on these measures on the first session compared to the last session to check for potential unwanted side-effects. STAI-s and s-AC were used to observe possible unwanted side-effects of NFT over time with a greater precision The questionnaires were programmed and presented using E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA).
5
study (N = 54). Frontal TBR was chosen as preselection outcome variable since all previous studies had frontal TBRas main outcome variable and this is highly correlated with central TBR (r = 0.902, p<0.001 in the current study).
We invited participants with the highest frontal TBR for the current study. Other inclusion criteria were: age between 18 and 24 years; no history of neurologic or psychiatric disorders; no history or current use of drugs other than low to moderate alcohol use or nicotine use and no use of medication that is known to directly influence the central nervous system. Recruitment took place at Leiden University, The Netherlands, between December 2015 and February 2016. All participants signed an informed consent and were free to terminate their participation at any time. For monetary compensation we used an incremental pay-off scheme, including disproportionately larger rewards for longer participation. This pay-off scheme was applied to minimize drop-out from the study. The study was approved by the local ethics review board (CEP16-011413), and pre-registered at ClinicalTrials.gov (NCT02763618).
Design
A single-blind case series multiple baseline design was used with a baseline (measurement-only period) varying prior to training onset and after training offset (Figure 5.1). Before the first lab session, the participants were assigned to one of three groups. Care was taken to obtain a more or less equal distribution of frontal TBR levels and age across the groups, but other than that the allocation to one of the three study groups was arbitrary. Allocation was done by the principal investigator who was not involved in the actual testing of the participants and had no contact with them. The experimenters that performed the study were kept blind to this allocation. All participants started with a three-session measurement-only phase with only a resting state EEG measurement. Participants in Group A continued with 14 sessions of active-NFT. Group B received six extra measurement-only sessions before they continued with eight active-NFT sessions. Group C received 14 sham-NFT sessions after the three-session measurement-only phase. A minimum of eight sessions active-NFT was applied because changes were usually seen around five or six 30-minute sessions in studies that found effects on theta frequency (Kao et al., 2014; Enriquez-Geppert et al., 2014). All participants were blind to which group they were in but the
experimenters were not blind to this for reasons of practical feasibility. During all sessions, questionnaires for state anxiety and state attentional control were assessed before every EEG measurement, active-NFT or sham-NFT. Our primary outcome variable was changes in frontal TBR within each session while our secondary outcome measurement was changes in frontal TBR between the sessions.
Figure 5.1. Difference in session course per group.
Materials
Self-Report Questionnaires.During the first and last session, participants completed the trait version of
the State-Trait Anxiety Inventory (STAI-Trait; Spielberger, 1983; van der Ploeg, Defares & Spielberger, 1980) and the Attentional Control Scale (ACS; Derryberry & Reed, 2002). The STAI-Trait assesses trait anxiety (20 items, range 20-80; Cronbach’s alpha = 0.89) and participants had to indicate their agreement with items like ‘I feel satisfied with myself’ and ‘I am a steady person’ on a four-point Likert scale. The ACS assesses self-reported attentional control in terms of attentional inhibition, attentional focus and the capacity to generate new thoughts (20 items, range 20-80; Cronbach’s alpha in present study = 0.85), e.g. ‘I can quickly switch from one task to another’. Self-reported state anxiety and state attentional control were measured on every session using the state version of the State-Trait Anxiety Inventory (STAI-State; Spielberger, 1983; van der Ploeg, Defares & Spielberger, 1980) and the
state-Attentional Control questionnaire (s-AC; Angelidis & Putman, manuscript in preparation). STAI- State measures
state anxiety at the moment of participation (20 items, range 20-80, Cronbach’s alpha = 0.91) and includes items like ‘I am tense’. s-AC measures attentional control at the moment of participation (6 items) and included items like ‘I feel very focused’ (Cronbach’s alpha = 0.84). The Behavioral Activation Scale (BAS; part of the Behavioral Inhibition and Activation Scale; BIS/BAS Carver & White, 1994) was assessed for the personality trait of behavioral activation (Cronbach’s alpha = 0.78). The BAS consists of the subscales BAS Reward, BAS Drive and BAS Fun Seeking but we assessed the total BAS score. STAI-t, ACS and BAS were included only to see if their scores changed on these measures on the first session compared to the last session to check for potential unwanted side-effects. STAI-s and s-AC were used to observe possible unwanted side-effects of NFT over time with a greater precision The questionnaires were programmed and presented using E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA).
EEG recording and Neurofeedback. The TBR Neurofeedback signal was measured and applied by a
on the jaw and right-ear mastoid respectively. Additionally, nine extra in-cap electrodes (BioSemi, The
Netherlands) were added on locations F3, Fz, F4, C3, Cz, C4, P3, Pz, P4, with one reference electrode on the left-ear mastoid, during every session. Data from these electrodes were collected with the Biosemi ActiveTwo DC amplifier. Both devices were active during all measurements, except during the sham-NFT sessions; then both the NeXus and the BioSemi system were attached but only the BioSemi system was active. Each session had a four-minute baseline measurement and a 25-four-minute measurement-only, active-NFT or sham-NFT. Active and sham neurofeedback were provided by BioTrace Software. Per time window of 15 seconds, individualized thresholds were automatically reset in a way that, based on the previous 15 seconds, the feedback signal would likely indicate successful performance for +/- 80% of the time, resulting in a standardized NFT protocol. Before feedback onset, measurement started 15 seconds earlier to determine the thresholds. When the EEG theta power went below the threshold, and the beta power above its threshold simultaneously, the participant was rewarded by the continuation of a video (a simulation of an airplane flying over a mountainous terrain). If they failed to reach these thresholds, the video stopped. Theta and beta amplitudes were online filtered with a 4 Hz high-pass and 7 Hz low-pass filter for theta and a 13 Hz high-pass and 30 Hz low-pass filter for beta. Online calculation of theta and beta amplitude (=feedback resolution) was done in epochs of 125 milliseconds using a moving time window; at every data sample (sampling rate was 256 per second) calculation was done over the last 125 milliseconds of that sample. These amplitude values were calculated by taking the root mean square (RMS) of the band-pass filtered data. Since online Fast Fourier transformation needs at least 2 seconds to calculate amplitudes; RMS is a representative and practical method of online calculation (Nitschke, Miller and Cook, 1998). Feedback by video continuation therefore appeared continuously when theta and beta were below and above the threshold for 125 milliseconds. The theta and beta amplitudes were visualized in separate bar graphs on the screen, next to the video. A third ‘inhibit’ bar represented voluntary eye blinks or muscle artifacts. The filter of this inhibit band was set at 2-3 Hz for eyeblinks and above 60 Hz for muscle artifacts. If the amplitude of the eyeblinks or muscle artifacts exceeded its threshold, no feedback was provided. The theta bar was a fluctuating bar in blue and beta a fluctuating bar in red. The participants were instructed which bar (beta) needed to go up above a threshold (small black stripe) and which bar (theta) needed to go down below a threshold, and in this way, they had to keep the video running. No instructions about how to influence their EEG spectrum were given. With respect to this, the participants were only told to ‘sit still’ and ‘not to tense their face or jaws’ (to reduce interference from muscle activity and to prevent increased beta activity resulting from such volitional motoric action).
The sham-NFT was a previously recorded active-NFT session of a participant from Group A (received 14 active-NFT sessions). Every participant in Group C (sham-NFT) was matched to another participant in group A and received the active-NFT video per session of their matched participant. That is, participants in Group C at session 4 saw the video and bars moving as if it was caused by their own EEG, however they actually watched the feedback that their matched participant from Group A in session 4 received. In this way, participants receiving sham-NFT were not able to influence the theta or beta bar graphs nor the continuation of the video. By matching every participant in Group C to a real participant in Group A, we kept the sham-NFT realistic for the participants in Group C, providing an accurate ‘yoked control’ procedure controlling for possible effects of motivation as a result
of the received feedback, for instance. Procedure
General procedure. Testing took place at Leiden University, between February 2016 until May 2016. All participants visited the lab 17 times. In all sessions, participants received state questionnaires, a four-minute EEG passive baseline resting state measurement followed by either a measurement-only, an active NFT or a sham-NFT (all 25 minutes; for an overview see Figure 5.1). Sessions were planned minimally three times a week with a maximum of five times a week. Every session took place on a separate day with a maximum of three days in between. One session approximately took between 60-70 minutes. The complete experiment therefore took 17-20 hours per participant in approximately four weeks. At the end of every session, all participants performed a 10-minute cognitive control task (Bishop, 2009). We included this task to pilot its extensively repeated use in a multiple baseline design for future studies. Results on this task are irrelevant for the current hypotheses and therefore the task and its outcomes will not be further reported.
First session. During the first session, participants were asked to sign an informed consent in which they were informed about the pay-off scheme regarding financial compensation. After signing the informed consent, a questionnaire about general and medical information was completed including questions about drug use and health. Participants started with the STAI-Trait, ACS, BAS, STAI-State and s-AC; in that order. Subsequently, preparation of the EEG equipment started and the participants continued with the four-minute passive baseline measurement followed by a 25 minute ‘measurement-only’ part.
Session 2 – 16. The second session till the sixteenth session, all maintained the same procedure, except that the fourth session till the sixteenth session could either include a measurement-only, an active NFT plus EEG measurement or sham-NFT plus EEG measurement (see Figure 5.1). Participants always started with completing the STAI-State and s-AC. This was followed by the EEG four-minute passive baseline measurement and the 25 minute measurement-only (session 2 and 3 for all groups and 2 – 9 for Group B), active-NFT (session 4 – 16 for Group A and session 10-16 for Group B) or sham-NFT (session 4-16 for Group C).
Last session (17). The last session started with completion of the STAI-Trait, BAS and trait ACS questionnaires, followed again by the STAI-State and s-AC questionnaires, EEG four-minute passive baseline measurement and the 25-minute active NFT or sham-NFT. All participants ended with a funneled debriefing interview. In this interview the participants were asked how they experienced the study, what kind of mental methods they used to become successful in the training, and which experimental group they thought they were in and why. After completing the interview, participants received a financial reward for their participation. Data Reduction and analysis
5
on the jaw and right-ear mastoid respectively. Additionally, nine extra in-cap electrodes (BioSemi, The
Netherlands) were added on locations F3, Fz, F4, C3, Cz, C4, P3, Pz, P4, with one reference electrode on the left-ear mastoid, during every session. Data from these electrodes were collected with the Biosemi ActiveTwo DC amplifier. Both devices were active during all measurements, except during the sham-NFT sessions; then both the NeXus and the BioSemi system were attached but only the BioSemi system was active. Each session had a four-minute baseline measurement and a 25-four-minute measurement-only, active-NFT or sham-NFT. Active and sham neurofeedback were provided by BioTrace Software. Per time window of 15 seconds, individualized thresholds were automatically reset in a way that, based on the previous 15 seconds, the feedback signal would likely indicate successful performance for +/- 80% of the time, resulting in a standardized NFT protocol. Before feedback onset, measurement started 15 seconds earlier to determine the thresholds. When the EEG theta power went below the threshold, and the beta power above its threshold simultaneously, the participant was rewarded by the continuation of a video (a simulation of an airplane flying over a mountainous terrain). If they failed to reach these thresholds, the video stopped. Theta and beta amplitudes were online filtered with a 4 Hz high-pass and 7 Hz low-pass filter for theta and a 13 Hz high-pass and 30 Hz low-pass filter for beta. Online calculation of theta and beta amplitude (=feedback resolution) was done in epochs of 125 milliseconds using a moving time window; at every data sample (sampling rate was 256 per second) calculation was done over the last 125 milliseconds of that sample. These amplitude values were calculated by taking the root mean square (RMS) of the band-pass filtered data. Since online Fast Fourier transformation needs at least 2 seconds to calculate amplitudes; RMS is a representative and practical method of online calculation (Nitschke, Miller and Cook, 1998). Feedback by video continuation therefore appeared continuously when theta and beta were below and above the threshold for 125 milliseconds. The theta and beta amplitudes were visualized in separate bar graphs on the screen, next to the video. A third ‘inhibit’ bar represented voluntary eye blinks or muscle artifacts. The filter of this inhibit band was set at 2-3 Hz for eyeblinks and above 60 Hz for muscle artifacts. If the amplitude of the eyeblinks or muscle artifacts exceeded its threshold, no feedback was provided. The theta bar was a fluctuating bar in blue and beta a fluctuating bar in red. The participants were instructed which bar (beta) needed to go up above a threshold (small black stripe) and which bar (theta) needed to go down below a threshold, and in this way, they had to keep the video running. No instructions about how to influence their EEG spectrum were given. With respect to this, the participants were only told to ‘sit still’ and ‘not to tense their face or jaws’ (to reduce interference from muscle activity and to prevent increased beta activity resulting from such volitional motoric action).
The sham-NFT was a previously recorded active-NFT session of a participant from Group A (received 14 active-NFT sessions). Every participant in Group C (sham-NFT) was matched to another participant in group A and received the active-NFT video per session of their matched participant. That is, participants in Group C at session 4 saw the video and bars moving as if it was caused by their own EEG, however they actually watched the feedback that their matched participant from Group A in session 4 received. In this way, participants receiving sham-NFT were not able to influence the theta or beta bar graphs nor the continuation of the video. By matching every participant in Group C to a real participant in Group A, we kept the sham-NFT realistic for the participants in Group C, providing an accurate ‘yoked control’ procedure controlling for possible effects of motivation as a result
of the received feedback, for instance. Procedure
General procedure. Testing took place at Leiden University, between February 2016 until May 2016. All participants visited the lab 17 times. In all sessions, participants received state questionnaires, a four-minute EEG passive baseline resting state measurement followed by either a measurement-only, an active NFT or a sham-NFT (all 25 minutes; for an overview see Figure 5.1). Sessions were planned minimally three times a week with a maximum of five times a week. Every session took place on a separate day with a maximum of three days in between. One session approximately took between 60-70 minutes. The complete experiment therefore took 17-20 hours per participant in approximately four weeks. At the end of every session, all participants performed a 10-minute cognitive control task (Bishop, 2009). We included this task to pilot its extensively repeated use in a multiple baseline design for future studies. Results on this task are irrelevant for the current hypotheses and therefore the task and its outcomes will not be further reported.
First session. During the first session, participants were asked to sign an informed consent in which they were informed about the pay-off scheme regarding financial compensation. After signing the informed consent, a questionnaire about general and medical information was completed including questions about drug use and health. Participants started with the STAI-Trait, ACS, BAS, STAI-State and s-AC; in that order. Subsequently, preparation of the EEG equipment started and the participants continued with the four-minute passive baseline measurement followed by a 25 minute ‘measurement-only’ part.
Session 2 – 16. The second session till the sixteenth session, all maintained the same procedure, except that the fourth session till the sixteenth session could either include a measurement-only, an active NFT plus EEG measurement or sham-NFT plus EEG measurement (see Figure 5.1). Participants always started with completing the STAI-State and s-AC. This was followed by the EEG four-minute passive baseline measurement and the 25 minute measurement-only (session 2 and 3 for all groups and 2 – 9 for Group B), active-NFT (session 4 – 16 for Group A and session 10-16 for Group B) or sham-NFT (session 4-16 for Group C).
Last session (17). The last session started with completion of the STAI-Trait, BAS and trait ACS questionnaires, followed again by the STAI-State and s-AC questionnaires, EEG four-minute passive baseline measurement and the 25-minute active NFT or sham-NFT. All participants ended with a funneled debriefing interview. In this interview the participants were asked how they experienced the study, what kind of mental methods they used to become successful in the training, and which experimental group they thought they were in and why. After completing the interview, participants received a financial reward for their participation. Data Reduction and analysis
high-pass filtered at 0.1 Hz, low-pass filtered at 100-Hz and a 50-Hz notch filter was applied. The data was automatically corrected for ocular artifacts (Gratton, Coles & Donchin, 1983) in segments of 4 seconds. Remaining segments containing muscle movements, amplitudes above 200 µV or other artifacts were removed. For offline amplitude calculation, Fast Fourier transformation (Hamming window length 10%) was applied for theta (4-7 Hz) and beta (13-30 Hz) at C3, Cz, and C4 positions. Amplitude values were calculated by taking amplitude spectral density (µV*Hz). Amplitude squared provided the power values. Central theta and beta power was calculated by taking the average of C3, Cz and C4 positions, and central TBR in turn was calculated by dividing central theta power by central beta power. Central TBR was chosen as outcome variable of focus because the NeXus sensor for active-NFT was placed between Cz and FCz positions, however we have exploratively looked at frontal average TBR too, as well as theta, beta, beta 1 (13-20 Hz) and beta 2 (21-30 Hz) separately. All raw data are freely available on (https://tinyurl.com/y948dcsl).
For interpreting the results, primarily visual inspection was used to determine the effectiveness of the active-NFT. If central TBR in Group A would reduce shortly after the introduction of the active-NFT (and after a comparable delay across the four participants) compared to no changes in Group C, the experiment would provide compelling evidence for the effectiveness of active-NFT. The effect would be even more strongly supported if a similar reduction was seen in Group A and B (after an equal number of active-NFT sessions, regardless the longer duration of the baseline). These changes after onset in Group A and Group B are assumed to be absent in Group C, where we expected no changes. The expected change in Group B could be considered as a direct replication of the effect in Group A. Furthermore, we expected differences in central TBR between Group A, B and C at the last session compared to the first session, with Group A showing the strongest reduction in central TBR (after performing the most active-NFT sessions), and Group C showing the weakest reduction in central TBR (no active NFT sessions). Primarily, we expect to see a consistent reduction in central TBR over active-NFT measurements, though we have inspected the four-minute passive baseline measurements as well, despite its smaller chance to detect any effect of active NFT. Also, besides inspecting changes in central TBR over all sessions, we have inspected changes in central TBR at the end of every session (average of last five minutes) and changes over time (25 minutes) within sessions too, since fluctuations might have occurred across the 25 minutes that could remain undetected when only inspecting session averages.
Trait anxiety and self-reported trait attentional control were measured at the start and the end of the study to exploratively relate these measures to possible changes in TBR as indication of potential unwanted side-effects. State anxiety and state-AC were assessed during every session to allow closer observation of such potential side effects.
Figure 5.2. Expected pattern of central TBR per group. Central TBR in Group A and B was expected to reduce some time after onset of active-NFT (session 4 for Group A and session 10 for Group B). The reduction is expected to be relatively constant between individuals. Central TBR in Group C was not expected to show any change over time because sham-NFT was introduced in session 4 instead of active-NFT. The thin black line represents the point of active-NFT introduction for Group A and Group B.
Results Participants
Twelve participants were selected, and all completed the 17 sessions (for a flow diagram of participant selection, see Figure 5.3). Age, baseline TBR and questionnaire scores per participant and per group during the first session are summarized in Table 3 The first baseline measurement of the selected 12 participants in the current study showed a frontal TBR of M = 1.51, SD = 0.76, median = 1.43). Although this was somewhat lower than their frontal TBR during their pre-selection measurement (M = 1.68, SD = 0.55, median = 1.47), it was still noticeably higher than the frontal TBR that was observed in the N=56 unselected sample that the preselection was based on (M = 1.26, SD = 0.54, median = 1.13) and represented the 45th-88th percentile score of this
5
high-pass filtered at 0.1 Hz, low-pass filtered at 100-Hz and a 50-Hz notch filter was applied. The data wasautomatically corrected for ocular artifacts (Gratton, Coles & Donchin, 1983) in segments of 4 seconds. Remaining segments containing muscle movements, amplitudes above 200 µV or other artifacts were removed. For offline amplitude calculation, Fast Fourier transformation (Hamming window length 10%) was applied for theta (4-7 Hz) and beta (13-30 Hz) at C3, Cz, and C4 positions. Amplitude values were calculated by taking amplitude spectral density (µV*Hz). Amplitude squared provided the power values. Central theta and beta power was calculated by taking the average of C3, Cz and C4 positions, and central TBR in turn was calculated by dividing central theta power by central beta power. Central TBR was chosen as outcome variable of focus because the NeXus sensor for active-NFT was placed between Cz and FCz positions, however we have exploratively looked at frontal average TBR too, as well as theta, beta, beta 1 (13-20 Hz) and beta 2 (21-30 Hz) separately. All raw data are freely available on (https://tinyurl.com/y948dcsl).
For interpreting the results, primarily visual inspection was used to determine the effectiveness of the active-NFT. If central TBR in Group A would reduce shortly after the introduction of the active-NFT (and after a comparable delay across the four participants) compared to no changes in Group C, the experiment would provide compelling evidence for the effectiveness of active-NFT. The effect would be even more strongly supported if a similar reduction was seen in Group A and B (after an equal number of active-NFT sessions, regardless the longer duration of the baseline). These changes after onset in Group A and Group B are assumed to be absent in Group C, where we expected no changes. The expected change in Group B could be considered as a direct replication of the effect in Group A. Furthermore, we expected differences in central TBR between Group A, B and C at the last session compared to the first session, with Group A showing the strongest reduction in central TBR (after performing the most active-NFT sessions), and Group C showing the weakest reduction in central TBR (no active NFT sessions). Primarily, we expect to see a consistent reduction in central TBR over active-NFT measurements, though we have inspected the four-minute passive baseline measurements as well, despite its smaller chance to detect any effect of active NFT. Also, besides inspecting changes in central TBR over all sessions, we have inspected changes in central TBR at the end of every session (average of last five minutes) and changes over time (25 minutes) within sessions too, since fluctuations might have occurred across the 25 minutes that could remain undetected when only inspecting session averages.
Trait anxiety and self-reported trait attentional control were measured at the start and the end of the study to exploratively relate these measures to possible changes in TBR as indication of potential unwanted side-effects. State anxiety and state-AC were assessed during every session to allow closer observation of such potential side effects.
Figure 5.2. Expected pattern of central TBR per group. Central TBR in Group A and B was expected to reduce
some time after onset of active-NFT (session 4 for Group A and session 10 for Group B). The reduction is expected to be relatively constant between individuals. Central TBR in Group C was not expected to show any change over time because sham-NFT was introduced in session 4 instead of active-NFT. The thin black line represents the point of active-NFT introduction for Group A and Group B.
Results Participants
Twelve participants were selected, and all completed the 17 sessions (for a flow diagram of participant selection, see Figure 5.3). Age, baseline TBR and questionnaire scores per participant and per group during the first session are summarized in Table 3 The first baseline measurement of the selected 12 participants in the
current study showed a frontal TBR of M = 1.51, SD = 0.76, median = 1.43). Although this was somewhat lower
than their frontal TBR during their pre-selection measurement (M = 1.68, SD = 0.55, median = 1.47), it was still
noticeably higher than the frontal TBR that was observed in the N=56 unselected sample that the preselection
was based on (M = 1.26, SD = 0.54, median = 1.13) and represented the 45th-88th percentile score of this
123 A g e Pr e-se le ct io n TB R A C S sc or e ST A I t ra it s co re BA S to ta l s co re G ro up A M = 2 0. 5 ± 1 M = 1 .7 4 ± 0 .8 1 M = 5 5. 5 ± 5 .9 7 M = 3 7 ± 2 .1 6 M = 4 3 ± 4 .5 4 Pa rt ic ip an t 2 21 1. 72 48 37 37 4 21 1. 19 58 36 42 5 19 2. 89 54 35 46 1 0 21 1. 15 62 40 47 G ro up B M = 2 1. 25 ± 1 .2 5 M = 1 .8 0 ± 0 .5 1 M = 5 0. 5 ± 1 1 M = 4 5. 25 ± 9 .7 8 M = 3 8 ± 7 .4 8 Pa rt ic ip an t 1 21 2. 50 45 36 36 7 21 1. 36 45 56 38 8 20 1. 88 45 51 30 1 1 23 1. 48 67 38 48 G ro up C M = 2 1. 25 ± 0 .9 6 M = 1 .5 ± 0 .3 2 M = 5 3. 75 ± 7 .8 9 M = 3 5 ± 4 .0 8 M = 4 4. 5 ± 3 .4 2 Pa rt ic ip an t 3 21 1. 93 50 38 44 6 20 1. 44 57 39 46 9 22 1. 16 63 32 40 1 2 22 1. 47 45 31 48 To ta l M ea n M = 2 1 ± 1 .0 4 M = 1 .5 1 ± 0. 55 M = 5 3. 25 ± 8 .0 2 M = 3 9. 10 ± 7 .3 1 M = 4 1. 83 ± 5 .7 0 Ta bl e 3. D emo gra ph ic in fo rma tio n o f a ll p art ic ip an ts a nd m ea ns p er g rou p. N ot e T BR = fron ta l t het a/ bet a r at io; M = m ea n, ± is st an da rd d ev ia tion .
5
Figure 5.4(ABC). Central TBR averages per 4 minute baseline measurement per participant (lines) in group A, B
and C. The thin black vertical lines represent when the active or sham NFT started.
Passive baseline between sessions
Each session started with a four-minute passive baseline measurement. Figure 5.4 shows the pattern of the average central TBR on the four-minute passive baseline per participant. A vertical line indicates the start of active-NFT or sham-NFT. We hypothesized that central TBR would reduce after the onset of active-NFT (in Group A and Group B) but would not show a consistent decrease or increase after the onset of sham-NFT (in Group C).
Visual inspection provided no support for such pattern. Passive baseline central TBR did not consistently change during the study, although some apparently random fluctuations between sessions were observed. This was invariably the case for all participants in both Groups A and B. No consistent increase or decrease of baseline central TBR was observed at any point in time in all participants. None of the participants that received sham-NFT showed a consistent decrease or increase after the onset of sham-NFT (Group C).
Average TBR during the active training phase of the sessions
5
Figure 5.4(ABC). Central TBR averages per 4 minute baseline measurement per participant (lines) in group A, B and C. The thin black vertical lines represent when the active or sham NFT started.
Passive baseline between sessions
Each session started with a four-minute passive baseline measurement. Figure 5.4 shows the pattern of the average central TBR on the four-minute passive baseline per participant. A vertical line indicates the start of active-NFT or sham-NFT. We hypothesized that central TBR would reduce after the onset of active-NFT (in Group A and Group B) but would not show a consistent decrease or increase after the onset of sham-NFT (in Group C).
Visual inspection provided no support for such pattern. Passive baseline central TBR did not consistently change during the study, although some apparently random fluctuations between sessions were observed. This was invariably the case for all participants in both Groups A and B. No consistent increase or decrease of baseline central TBR was observed at any point in time in all participants. None of the participants that received sham-NFT showed a consistent decrease or increase after the onset of sham-NFT (Group C).
Average TBR during the active training phase of the sessions
Figure 5.5(ABC). Central TBR averages during the 25-minute sessions of measurement only or active NFT or sham NFT training for Groups A, B and C.
Last five minutes of training phase
The possibility exists that calculating an average over a longer period of time will obscure any delayed within-session effects of active-NFT. In other words, active-NFT might reduce central TBR at the end of sessions only, for example because the learning-process takes time. To check for this possibility, we explored changes between sessions in average central TBR during the last five minutes of every session. Figure 5.6A, B, and C show
the pattern of the central TBR over sessions per participant of the averaged final 5 minutes. No clear differences between measurement-only, active-NFT and sham-NFT sessions became visible.
Figure 5.6(ABC). Central TBR averages of last 5 minutes per measurement-only/active/sham NFT session per
5
Figure 5.5(ABC). Central TBR averages during the 25-minute sessions of measurement only or active NFT or sham
NFT training for Groups A, B and C.
Last five minutes of training phase
The possibility exists that calculating an average over a longer period of time will obscure any delayed within-session effects of active-NFT. In other words, active-NFT might reduce central TBR at the end of sessions only, for example because the learning-process takes time. To check for this possibility, we explored changes between sessions in average central TBR during the last five minutes of every session. Figure 5.6A, B, and C show
the pattern of the central TBR over sessions per participant of the averaged final 5 minutes. No clear differences between measurement-only, active-NFT and sham-NFT sessions became visible.
Figure 5.6(ABC). Central TBR averages of last 5 minutes per measurement-only/active/sham NFT session per
Chapter 5
Inspection per case
Finally, we examined whether any consistent change in central TBR could be detected within the sessions. We calculated the average central TBR per minute for every session and plotted these over time (25 minutes on the x-axis) for each session and participant. All plots are available online (https://tinyurl.com/y948dcsl). Developments of central TBR over time were compared within sessions between measurement-only, active-NFT and sham NFT conditions. The effect of active-NFT might for example have been driven by motivation, or inhibited by fatigue, factors that are likely a function of the duration of a session and that are different for each session and participant. Visual inspection did not reveal any clear change over central TBR within the active-NFT sessions. We here present some detailed data for one of the ‘best cases’, showing some kind of change in central TBR/active-NFT effect. This concerns two out of four participants in Group B, in whom central TBR seemed to decrease within some active-NFT sessions compared to the measurement-only sessions. Here, we present these detailed data only for participant 7 (Figure 5.7C and 7D. See https://tinyurl.com/y948dcsl for all data of all participants). The reduction in central TBR occurred in sessions 10 and 13 but not anymore in sessions 15 or 17. Furthermore, central TBR always started at approximately the same value in all active NFT sessions of participant 7. The data therefore do not show any transfer of a learning process caused by the active-NFT intervention.
Figure 5.7(ABCDEF). Central TBR over time for participant 7 within; session 1 (A) session 9 as the last
measurement-only session (B), session 10 as the first active NFT session (C), session 13 and 15 as two in between active NFT sessions (D and E) and session 17 as the last session (F).
State anxiety and state AC over time
To check for potential adverse effects, scores on STAI-s and s-AC were plotted for all participants over all
129
sessions. We calculated the average central TBR per minute for every session and plotted these over time (25 minutes on the x-axis) for each session and participant. All plots are available online (https://tinyurl.com/y948dcsl). Developments of central TBR over time were compared within sessions between measurement-only, active-NFT and sham NFT conditions. The effect of active-NFT might for example have been driven by motivation, or inhibited by fatigue, factors that are likely a function of the duration of a session and that are different for each session and participant. Visual inspection did not reveal any clear change over central TBR within the active-NFT sessions. We here present some detailed data for one of the ‘best cases’, showing some kind of change in central TBR/active-NFT effect. This concerns two out of four participants in Group B, in whom central TBR seemed to decrease within some active-NFT sessions compared to the measurement-only sessions. Here, we present these detailed data only for participant 7 (Figure 5.7C and 7D. See https://tinyurl.com/y948dcsl for all data of all participants). The reduction in central TBR occurred in sessions 10 and 13 but not anymore in sessions 15 or 17. Furthermore, central TBR always started at approximately the same value in all active NFT sessions of participant 7. The data therefore do not show any transfer of a learning process caused by the active-NFT intervention.
Figure 5.7(ABCDEF). Central TBR over time for participant 7 within; session 1 (A) session 9 as the last
measurement-only session (B), session 10 as the first active NFT session (C), session 13 and 15 as two in between active NFT sessions (D and E) and session 17 as the last session (F).
State anxiety and state AC over time
To check for potential adverse effects, scores on STAI-s and s-AC were plotted for all participants over all
129
Inspection per case
Finally, we examined whether any consistent change in central TBR could be detected within the sessions. We calculated the average central TBR per minute for every session and plotted these over time (25 minutes on the x-axis) for each session and participant. All plots are available online (https://tinyurl.com/y948dcsl). Developments of central TBR over time were compared within sessions between measurement-only, active-NFT and sham NFT conditions. The effect of active-NFT might for example have been driven by motivation, or inhibited by fatigue, factors that are likely a function of the duration of a session and that are different for each session and participant. Visual inspection did not reveal any clear change over central TBR within the active-NFT sessions. We here present some detailed data for one of the ‘best cases’, showing some kind of change in central TBR/active-NFT effect. This concerns two out of four participants in Group B, in whom central TBR seemed to decrease within some active-NFT sessions compared to the measurement-only sessions. Here, we present these detailed data only for participant 7 (Figure 5.7C and 7D. See https://tinyurl.com/y948dcsl for all data of all participants). The reduction in central TBR occurred in sessions 10 and 13 but not anymore in sessions 15 or 17. Furthermore, central TBR always started at approximately the same value in all active NFT sessions of participant 7. The data therefore do not show any transfer of a learning process caused by the active-NFT intervention.
Figure 5.7(ABCDEF). Central TBR over time for participant 7 within; session 1 (A) session 9 as the last
measurement-only session (B), session 10 as the first active NFT session (C), session 13 and 15 as two in between active NFT sessions (D and E) and session 17 as the last session (F).
State anxiety and state AC over time
To check for potential adverse effects, scores on STAI-s and s-AC were plotted for all participants over all
EEG Theta/Beta Ratio Neurofeedback Training in Healthy Females.
5
Inspection per caseFinally, we examined whether any consistent change in central TBR could be detected within the sessions. We calculated the average central TBR per minute for every session and plotted these over time (25 minutes on the x-axis) for each session and participant. All plots are available online (https://tinyurl.com/y948dcsl). Developments of central TBR over time were compared within sessions between measurement-only, active-NFT and sham NFT conditions. The effect of active-NFT might for example have been driven by motivation, or inhibited by fatigue, factors that are likely a function of the duration of a session and that are different for each session and participant. Visual inspection did not reveal any clear change over central TBR within the active-NFT sessions. We here present some detailed data for one of the ‘best cases’, showing some kind of change in central TBR/active-NFT effect. This concerns two out of four participants in Group B, in whom central TBR seemed to decrease within some active-NFT sessions compared to the measurement-only sessions. Here, we present these detailed data only for participant 7 (Figure 5.7C and 7D. See https://tinyurl.com/y948dcsl for all data of all participants). The reduction in central TBR occurred in sessions 10 and 13 but not anymore in sessions 15 or 17. Furthermore, central TBR always started at approximately the same value in all active NFT sessions of participant 7. The data therefore do not show any transfer of a learning process caused by the active-NFT intervention.
Figure 5.7(ABCDEF). Central TBR over time for participant 7 within; session 1 (A) session 9 as the last
measurement-only session (B), session 10 as the first active NFT session (C), session 13 and 15 as two in between active NFT sessions (D and E) and session 17 as the last session (F).
State anxiety and state AC over time
To check for potential adverse effects, scores on STAI-s and s-AC were plotted for all participants over all
129
sessions. We calculated the average central TBR per minute for every session and plotted these over time (25 minutes on the x-axis) for each session and participant. All plots are available online (https://tinyurl.com/y948dcsl). Developments of central TBR over time were compared within sessions between measurement-only, active-NFT and sham NFT conditions. The effect of active-NFT might for example have been driven by motivation, or inhibited by fatigue, factors that are likely a function of the duration of a session and that are different for each session and participant. Visual inspection did not reveal any clear change over central TBR within the active-NFT sessions. We here present some detailed data for one of the ‘best cases’, showing some kind of change in central TBR/active-NFT effect. This concerns two out of four participants in Group B, in whom central TBR seemed to decrease within some active-NFT sessions compared to the measurement-only sessions. Here, we present these detailed data only for participant 7 (Figure 5.7C and 7D. See https://tinyurl.com/y948dcsl for all data of all participants). The reduction in central TBR occurred in sessions 10 and 13 but not anymore in sessions 15 or 17. Furthermore, central TBR always started at approximately the same value in all active NFT sessions of participant 7. The data therefore do not show any transfer of a learning process caused by the active-NFT intervention.
Figure 5.7(ABCDEF). Central TBR over time for participant 7 within; session 1 (A) session 9 as the last
measurement-only session (B), session 10 as the first active NFT session (C), session 13 and 15 as two in between active NFT sessions (D and E) and session 17 as the last session (F).
State anxiety and state AC over time
To check for potential adverse effects, scores on STAI-s and s-AC were plotted for all participants over all
Figure 5.8(AB). Scores on state anxiety (A) and state attentional control (B) per session per participant in Group A, B or C.
Motivation and debriefing
All participants generally reported to be motivated performing the NFT over all sessions (M = 3.35; range
1-4 with 4 being most motivated), although for all three Groups (A, B and C) there was a small drop in motivation
between the 10th and the 12th session. Motivation returned to their initial level after the 12th session. All
participants received a funneled debriefing interview after the final session of the study, and it became clear that two out of the four participants that received NFT (Group C) were not sure whether they were in the
sham-controlled group (60% chance of being in the sham sham-controlled group) whilst the two other participants in Group C thought that they had received an active-NFT (70% chance).
Discussion
In this study we aimed to reduce central TBR with NFT in healthy individuals with elevated TBR. The results indicate that active-NFT did not alter TBR in any way. No consistent within-session change of TBR was found on either the passive baseline measurement or 25-minute active training measurement nor was there any evidence of between-session change. This suggests that the active-NFT did not induce any changes in EEG measurements of interest. State anxiety and state attentional control did not show a consistent change after active-NFT onset either. All participants reported to remain motivated performing the active- or sham-NFT however, and participants that received sham-NFT indicated that they believed to have received an active-NFT.
The present study was a first step towards intervention studies in a healthy population with elevated TBR. We expected that NFT would reduce TBR in healthy participants. Changes in EEG were the primary outcome, as these changes are likely easier to detect and more consistent than changes at the more multifaceted and complex behavioral level. We used a multiple baseline case series design for a detailed study of all NFT effects.
Our finding that active-NFT did not induce any consistent reduction or increase in TBR in healthy individuals is in line with the results of Doppelmayr and Weber (2011) who performed a randomized controlled trial in 14 healthy participants. Thirty active-NFT sessions did not induce changes in EEG TBR or the separate theta or beta frequency bands. Their results do not provide explanation why the active-NFT did not alter TBR. Possibly, some changes were simply not detected because TBR changes were not inspected within-sessions. By using a multiple baseline case series design, we provide a detailed view of what precisely happened with EEG TBR over time after the onset of active-NFT over sessions, as well as a precise view of TBR changes within active NFT sessions, over the course of 25 minutes. None of our detailed observations provided any evidence that active-NFT had an effect on the primary outcome variable, the EEG spectrum level of TBR, replicating the results of
Doppelmayr and Weber (2011).
